mechanistic investigation of timber crosstie spike failures · 2020-07-07 · railtec at illinois |...

Mechanistic Investigation of Timber Crosstie Spike Failures

Matheus TrizottoChristian KhachaturianMarcus DerschJ. Riley EdwardsArthur LimaJune 04

Industry Partners Update Meeting

RailTEC at Illinois | 2

Acknowledgements

► Project Sponsor

► Industry Partnership


Outline

► Problem Overview► Project Summary

• Field, Lab, and Modelling► Wheel and Rail Seat Loads – Field Results

• Vertical• Lateral• Longitudinal

► Spike Stress Mitigation Methods - Modeling• Cross Sectional Area• Screw vs Cut• Load Location• Plate Clamping Force• Spike Engagement

► TTC Results• Test vs Control

► Preliminary Recommendations for Mitigating Spike Failures


Traditional System with Anchor

Background on Spike Failures

► Spikes have been failing due to fatigue• Primarily on premium fastening system• Often new ties• Mainly on curves• Also in special trackwork

► Driven by overloading mechanism

Dersch et. al. 2018

Longitudinal Rail Seat Load

Plate

Rail Anchor Spikes Plate-Tie

Friction

Crosstie

Rail-Plate Friction


Traditional System with Anchor



► Driven by overloading mechanism► Leads to rapid gauge deterioration causing

at least 10 derailments since 2000

Dersch et. al. 2018

Likely Driving Factors► Lack of anchor increases longitudinal loads


Plate

Rail Anchor Spikes Plate-Tie

Friction

Crosstie

Rail-Plate Friction

Mosier, OR - 2016 (Sean Aiken)


Premium System with Clip





Dersch et. al. 2018


• Wood is weaker in this direction


Spikes

Crosstie

Elastic Clip

Plate

Rail-Plate Friction

Plate-Tie Friction







Dersch et. al. 2018


• Wood is weaker in this direction► Plate uplift further increases loads► Stiffer fasteners reduce distribution of loads► Crosstie age


Spikes

Crosstie

Plate

Elastic Clip

Premium System with Clip

Rail-Plate Friction



Lateral

LateralLateral

LateralLateral

LongitudinalLongitudinal

Stress in Spikes – Hypothetical Graph

AnchorCurveGrade

AnchorCurveGrade

(extreme case)

Total Stressinto Spikes

Threshold Stress for Spike Failure

AnchorTangent

Flat

Longitudinal

Longitudinal

No AnchorCurveGrade

Longitudinal

AnchorCurveGrade

Traditional Systems Premium Systems


Project Thrust Areas

Laboratory Analytical Modeling Field

Controlled experiments to better characterize load transfer and rail displacement/slip of various fasteners

Quantification of load distribution and spike stress while investigating key factors affecting load transfer

Quantification of loading demands (vert., lateral, long.) and fastener response to service loading


Thrust Area Interaction

Laboratory

Analytical ModelingField

A. C.

B.

D.

A. Lab Field:• Field data can inform lab

loading magnitude• Controlled experiments

can challenge/replicate field data and inform future field work

B. Field Model:• Field provides load input

for model

C. Lab Model: • Lab provides data for

calibration of models• Model helps to focus lab

experiments

D. Provides opportunity to make comprehensive recommendations


Field Experimentation Motivation

Data acquisition through field experimentation allows for

► Quantification of Loading Environment• Improve our understanding of the load demands placed on the

fastening systems as a result of passing trains

► Evaluation of Fastening System Response• Improve our understanding of the characteristic stiffnesses,

deformations, and displacements as demands/conditions (loading, weather, etc.)

► Development of Analytical Model• Validate a three dimensional (3D) finite element (FE) model • Compare data with laboratory results

High Level Outcomes


Field Experimentation OverviewHorseshoe Curve

► 3-track curve in Norfolk Southern’s Pittsburgh line► Westbound track has primarily uphill empty trains (45.6 MGT)

● Tractive effort is distributed throughout the locomotives► Eastbound track has primarily downhill loaded trains (50 MGT)

● (Air)break forces are distributed throughout the entire train

Key feature: Tracks have the same curvature, grade and climate, allowing the comparison of the effects of the following on loading demand and fastener response

• Anchor• Temperature• Load• Speed

• Direction of traffic• Braking vs.

traction forces• High vs. low rail

Photo:Wikipedia


Quantification of Loading Environment► Vertical, lateral and longitudinal loads collected with industry

standard load circuits installed in the center of the crib on both high and low rails

n.a.

Vertical Loads

Lateral Loads

Longitudinal Loads


Downhill

Track 1(50 MGT)

Few Broken Spikes

Field Experimentation ReviewTrack 2 Track 3

(45.6 MGT)

Instrumentation Area

Data Collection Box

Wheel Counter

Uphill

Grade: 1.76%Curvature: 9.2°

LubricatedAirbrake grade

Many Broken Spikes


Track OverviewTrack 2 Track 3

(45.6 MGT)

Instrumentation Area

Data Collection Box

Wheel Counter

UphillDownhill

Grade: 1.76%Curvature: 9.2°

LubricatedAirbrake grade

Many Broken Spikes

Track 1(50 MGT)

Few Broken Spikes


Percent Exceeding Vertical Wheel Loads

► Vertical wheel loading is generally greater in Track 1• Expected, since operations of loaded trains are biased on that track

► Track 3 operations are seen to be underbalanced• In average, low rail vertical loads are more than 30% greater than in high rail

(July 2019) (43% of Wheel Load)


Percent Exceeding Vertical Wheel Loads(July 2019)

Verti

cal W

heel

Loa

d (k

ips)

Verti

cal R

ail S

eat L

oad

(kip

s)

Spee

d (m

ph)

► Vertical wheel loading is generally greater in Track 1• Expected, since operations of loaded trains are biased on that track

► Track 3 operations are seen to be underbalanced• In average, low rail vertical loads are more than 30% greater than in high rail


Percent Exceeding Lateral Wheel Loads

► Track 1 field-facing lateral wheel loads are measured to be, on average, 130% higher than in Track 3

► 10% of lateral loads are gauge-facing in Track 1 and 30% in Track 3

(July 2019)(60% of Wheel Load)


Percent Exceeding Lateral Wheel Loads(July 2019)

High Rail Low Rail-20

-10

0

10

20

30La

tera

l Whe

el L

oad

(kip

s)Track 1

n = 17681

Late

ral R

ail S

eat L

oad

(kip

s)(6

0% o

f Whe

el L

oad)

► Track 1 field-facing lateral wheel loads are measured to be, on average, 130% higher than in Track 3

► 10% of lateral loads are gauge-facing in Track 1 and 30% in Track 3


/

𝑘𝑘

Longitudinal LoadingWheel, rail axial, and fastener

Axial Load

Longitudinal Fastener Load

50%

𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒

0RailAxial Load

Following elastic track model by Kerr(2003)


Longitudinal Loading

/

𝑘𝑘

Axial Load

Longitudinal Fastener Load

ka/2 300 lbs/in/in

𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒0

Wheel, rail axial, and fastener

RailSeat Load

Following elastic track model by Kerr(2003)


Superposition of Axial Loads

RailAxial Load

► Cancellation of rail axial forces occur between locomotives

Locomotive Locomotive

Distribution of verticaland lateral forces

Distribution of longitudinal forces


Superposition of Fastener Loads

► Fastener loading is greatest between locomotives► Plate uplift can occur during application of longitudinal loading► Load superposition studied:

• Draft paper: “Analytical Elastic Modelling of Railroad Track Response”• Initial feedback collected from reviewers

Locomotive Locomotive

~Plate uplift zones

RailSeat Load


Rail Seat Load Example Calculation

Rail seat force estimation example:► Choose datapoints (two in this example)► Create a generic function for axial loads based on number of

locomotives► Solve for longitudinal wheel loads and track stiffness► Generate function for rail seat forces for the same number of

locomotives and find maximum

𝑅 9 𝑘𝑖𝑝𝑠𝑘 1732 𝑙𝑏𝑠/𝑖𝑛/𝑖𝑛


Rail Seat Load Example Calculation

𝐹 1.1 𝑘𝑖𝑝𝑠

𝑅 9 𝑘𝑖𝑝𝑠𝑘 1732 𝑙𝑏𝑠/𝑖𝑛/𝑖𝑛


Effect of Distributed Power on Fastener Loads

► Increased distance between trucks reduce superposition effects► This example shows a 40% reduction of fastener demand by

having two locomotives just 1 car apart

Locomotive LocomotiveBuffer

RailSeat Load


Measured Rail Axial Forces

► The average magnitude of measured rail axial forces are 50-100% higher on Track 3

► An analysis of load distribution was necessary to understand if longitudinal rail seat loading is greater on Track 3

(July 2019)Lo

ngitu

dina

l Rai

l Axi

al L

oad

(kip

s)


Estimated Longitudinal Fastener Forces

► Maximum longitudinal rail seat loads are estimated to be 10-20x lower than rail axial forces

► Longitudinal fastener loading on Track 3 predicted to be 2-5x higher than Track 1 (in average)• Reasonable considering differences between traction and braking

(July 2019)

Track stiffness and longitudinal wheel forces estimated and used for calculation of rail seat forces

Estim

ated

Lon

gitu

dina

l Rai

l Sea

t Loa

d (k

ips)


Estimated Longitudinal Fastener Forces

► 95th percentile longitudinal loads are 2-3x greater on Track 3 compared to Track 1► 95th percentile lateral loads are 2-2.5x greater on Track 1 compared to Track 3► If these loads were transferred to spikes, fatigue failures would occur in all rails

Estimated Longitudinal Fastener Loads (kips)Track 1 Track 3

Percentile High Rail Low Rail High Rail Low Rail99 0.92 0.60 1.87 2.5995 0.87 0.54 1.55 1.7990 0.78 0.50 1.42 1.7250 0.41 0.25 0.94 1.21

Lateral Fastener Loads (60% of rail load) (kips)Track 1 Track 3



Field Conclusions and Path Forward

► Longitudinal rail seat loads (up to ~2,600 lbs.) are one order of magnitude lower than rail axial load (up to ~25,000 lbs.)

► Underbalanced speed operations on Track 3 lead to 30% greater vertical loads on low rail compared to high rail

► Load demand itself does not explain why failure is predominant on the high rail of Track 3

Future Work:► Investigate the effect of operational characteristics and friction on rail

seat load transfer to crossties• Longitudinal

- Loads are applied before uplift occurs - Uplift likely mitigates the effect of friction on longitudinal load transfer

• Lateral- Loads applied roughly with vertical loads- Friction should help transfer lateral loads


Hypothetical: Rail Seat Load ComponentsLateral (Spike), Lateral Friction, and Longitudinal (Spike)

Coefficient of friction assumed to be 0.20 between plate and crosstie

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

High Rail Low Rail High Rail Low Rail

Track 1 Track 3

Late

ral +

Lon

gitu

dina

l Loa

d (k

ips)

Longitudinal (Spike) Longitudinal (Friction) Lateral (Spike) Lateral (Friction)


Single Spike FEM Summary

► Longitudinal load is more detrimental than an equivalent lateral load• This is due to the timber being less resistant in that direction• This finding supports the theory that spike failures in premium fastening

systems are primarily related to longitudinal loads► Stress can exceed the fatigue limit of spike steel at regular service loads► Max. stress depth varies with the magnitude and direction of applied load► Species of timber significantly effects the depth/load of failure

RailTEC at Illinois


Recent FEA Investigation Focus: Spike Stress Mitigation Methods► Spike cross-sectional area ► Spike type (cut vs screw)► Spike loading location► Plate engagement with tie (vertical clamping force)► Spike engagement with plate ► Quantity of spikes within a given fastening system





Modeling Overview




► Three basic models were developed for this study:• Single cut spike-timber block model

- Quantify effect of cross-sectional area and loading location• Single screw spike-timber block model

- Quantify effect of spike type• Single rail seat model with multiple spikes, plate, and timber block

- Quantify effect of spike/plate/tie engagement and quantity of spikes used


► At least one manufacturer has discussed using larger spikes with larger cross-sectional area

► Stress (σ) is a function of the moment (M), area moment of inertia (I) and distance from the neutral axis (y)

where, and,

► Increasing cross-sectional area would increase I at a greater rate than y

► Stress would be lower for a given moment (i.e. applied load)

Cross-Sectional Area: Background

𝜎𝑀𝑦𝐼 𝐼

𝑏ℎ12

𝑦𝑏2


► Each spike subjected to 2,500 lb. longitudinal load

► Inverse linear relationship between cross-sectional area and stress► For each 1% increase in standard spike width there is ~ 2% reduction in stress► Increasing spike width by 0.1” resulted in 29% reduction in stress

Effect of Cross-Sectional Area


► Multiple railroads are moving from cut spikes to screw spikes for premium fasteners• At least one railroad has moved from screw spikes to cut spikes

► Perceived advantages of screw spikes (relative to cut spikes):• Increased bending strength• Increased plate hold-down strength

► Expectations based on geometry (relative to cut spikes):• Increased bending strength at the upper shaft• Decreased bending strength at threads• Increased stresses in the threads at equivalent loads

Cut vs Screw Spikes: Background



► Maximum spike stress depth:• Cut spike: 1.5”• Screw spike: 2.0”

Cut vs Screw Spike: Visualization



► Longitudinal loads < 2000 lb. there is < 10% difference in max stress► Beyond 2000 lb. stress in screw spikes increase rapidly

Cut vs Screw Spike: Quantification


► Currently it is assumed that the load transfer into the spikes is uniform► Reasons for non-uniform are as follows:

• (a) the angle of spike being driven into the tie• (b) the non-planar finished surface of the spike or plate• (c) plate uplift

► To quantify the effect of plate uplift, the location of applied load will be varied

Loading Location: Background



► Direct linear relationship between change in load location and max stress► +/- 20% change from control case when load moves up/down, relatively

Effects of Loading Location


Normal Clamping Force: Background




Goal:► Ensure load is transferred through spike

and friction when plate is engaged with tieChallenge:► Wave action of the rail leads to uplift of

the plate, eliminating frictionProposed Solution:► Develop friction force through

preloading of plate:• Screw spikes do this upon first

installation, but can loosen over time• Spring washers could be utilized

(and are utilized internationally)► Spring washer applied load studied:

• 3,400 lb. ea. after initial “settling”• 1,000 lb. ea. after minimal cutting/loosening Load vs deflection of spring washer

3,400 lb.

1,000 lb.

~10,000 lb.*

* Example of load vs disp. for this spring washer, not an indication of load at installation.


Effect of Normal Clamping Force




► Fastening system subjected to 2,500 lb. longitudinal load ► Normal clamping force of 1,000lb./spike and 3,400lb./spike investigated

► 3,370lb./spike resulted in an 80% reduction in spike stress (on average)► 1,000lb./spike resulted in a 70% reduction in spike stress (on average)► Thus, most of longitudinal force taken by friction


Spike Engagement and Quantity of Spikes: Background

► Plate hole tolerances are larger than potential displacement• Hold down and line spike holes are 0.065 in. and 0.125 in. larger than the

0.625 in. standard cut spike• Displacement of spike subjected to load not expected to exceed 0.04 in.

► Not all spikes will be in contact at one time► Many railroads use 4 spikes within premium fasteners

• Some use 5 in most demanding locations/after maintenance

spike

hole

engageddisengaged


Effect of Spike Engagement with Plate




► Fastening system with 4 spikes subjected to 2,500 lb. longitudinal load with one spike moved to other side of spike hole (extreme tolerance)

► Disengaged spike acts as though it has been removed from system► Load is not evenly redistributed among engaged spikes► Disengagement of a single spike resulted in up to a 140% (2.4 times)

increase in spike stress


Effect of Quantity of Spikes




► Fastening system with 3 - 5 spikes subjected to 2,500 lb. longitudinal load

► Increasing spike quantity will not ensure max stress reduction for all spikes► Addition of fifth spike reduced stress of spikes 3 and 4, not 1 and 2


Modelling Conclusions

1. The disengagement of a single can lead to a 140% (2.4 times) increase in a remaining spike’s stress

2. There is an inverse relationship between plate-sleeper normal clamping force and spikes stress• Clamping force of 1,000 lb./spike can reduce spike stress by 70%

3. There is little difference in cut and screw spike strength (capacity) at longitudinal loads below 2,000 lbs. (8.90 kN)

4. There is an inverse linear relationship between spike cross-section and resulting stress• For each 1% increase in spike width there is ~ 2% reduction in stress• Increasing spike width by 0.1” resulted in 29% reduction in stress

5. There is direct linear relationship between spike contact location and resulting stress• A +/- 0.245” change in load location can lead to -/+ 20% change in

maximum stress► Draft journal paper submitted to JRRT

• “Methods to Mitigate Railway Premium Fastening System Spike Fatigue Failures using Finite Element Analysis”


TTC Field Work Overview ► Control zone:

• Victor Plates with e-clips and cut spikes• 20 crossties

► Test zone:• Vossloh fastener with screw spike, spiral

washer, tension clamp• 30 crossties

► Installed: Fall 2019

► Tonnage Accumulated: 170 MGT

► Visual Inspections: performed every few months

► Track Geometry and Loaded Gage: measured after fall and spring seasons


► Details of “loose screw spikes” within Test zone (after 60 MGT)• 3 of 240 exhibited 30 – 45 degrees of rotation (1 – 1.5 mm of uplift)• Spring washers estimated to still apply between 2 – 3 kips of clamping force

► Indications of additional “loose screw spikes” after 170 MGT on 5 rail seats

Control TestPlate Cutting minimal none visibleBroken Components 6 broken spikes none visibleLoosened Components spike uplift 3 (1.3%) screw spikes (at 60 MGT)Plate Movement some evident none visibleSkewed Crossties several ties (reset multiple times) two ties, minimal (after 140 MGT)

TTC Visual Inspection Results

Tie skewing example Potential spike uplift

Broken spike


TTC Field Gage Results and Discussion

► Spring washers encourage plate to crosstie contact• Increases friction between

plate and crosstie• Finite element analysis

indicates a benefit• Indicated in reduced base

gage widening► Tension clamps provide

improved/consistent clamping force compared to e-clip• More consistent toe-load• Increased lateral/rotational

strength/stiffness• Indicated in reduced head

gage widening

Potential factors leading to improved performance of Test Zone


Preliminary Recommendations for Mitigating Spike Failures

► Through Operations:• Reduce longitudinal demands by:

- Distributing power where possible/feasible- Evenly distributing tractive effort among multiple units

• Reduce lateral forces when possible by:- Proper gauging (wide gauge), - TOR lubrication, - etc.

• Operating near balanced speed to encourage equal vertical loading on high and low rail

• Building trains limiting unloading of vertical rail (string line condition, etc.)

► Through Fastening System Design:• Encourage engagement (clamping/contact) between plate and crosstie• Encourage engagement of all spikes with plate• Increase strength of spike(s) to account for additional loading


Acknowledgements

► Project Sponsor:

► Industry Partnership:


Thank you for your attention!

University of Illinois at Urbana-Champaign (UIUC)Rail Transportation and Engineering Center (RailTEC)

This project is supported by the National University Rail Center (NURail), a US DOT-OST Tier 1 University Transportation Center

Graduate Research [email protected]

Matheus TrizottoPrincipal Research Engineer

[email protected]

Marcus DerschGraduate Research Assistant

[email protected]

Christian Khachaturian


APPENDIX


Vertical Wheel Loads (July 2019)

Vertical Wheel Loads (kips)

Track 1 Track 3



Rail Axial Loads (Constrained Rail)Axial Loading Mechanisms► Uniform temperature changes

• Rail tries to expand in the heat – Compressive Loads• Rail tries to contract in the cold – Tensile Loads

- No relative rail movement → no fastener loading► Temperature gradient► Train tractive/ breaking efforts

• Produce directional movement- Therefore introducing fastener reaction

• Today’s focus

Constrained

Free bar

𝜎

𝑡𝑖𝑚𝑒


Axial Loading Profile

Track 1: Downhill Trains• Every axle is contributing

to braking• Uniform forces considered

Track 3: Uphill Trains• Only locomotives are

introducing tractive effort• Uniform forces considered

Nloco

1

2

3

4

1

2

3

4


Axial Loading Profile

Track 1: Downhill Trains• Every axle is contributing

to braking• Uniform forces considered

Track 3: Uphill Trains• Only locomotives are

introducing tractive effort• Uniform forces considered

Signature axial loads differ


Estimated Average Long. Wheel Load (R)

► Calculated wheel loads indicate that tractive effort in track 3 is, in average, +80% higher than breaking effort in track 1

► However, longitudinal loading in track 1 occurs in a longer section, and an increased number of wheels should be considered in the superposition of fastener forces


Estimated Longitudinal Track Stiffness (k)

► Stiffness ranges are in the order of magnitude of shakedown laboratory tests

► The methodology estimates Track 3 to be 2-4x stiffer than Track 1 *

Estim

ated

Lon

gitu

dina

l Stif

fnes

s (lb

s/in

/in)

(July 2019)


RailSeat Load

Track Response: Change in Stiffness

► Higher stiffness leads to increased fastener forces, but also reducesload superposition

► * Longitudinal track stiffness are not the greatest factor that affectsrail seat loads• Reducing the longitudinal stiffness to a quarter (from 4000 to 1000 lbs/in/in)

reduced maximum fastener loads by 25%


0

5

10

15

20

25

30

35

40

45

0 5 10 15 20 25 30

Load

(kip

s)

Speed (mph)

Load vs Speed on High and Low Rail

Low RailHigh Rail

Mea

sure

d Sp

eed


Lateral Wheel Loads By Axle Position

► Lateral wheel loads in lead axles observed to be 30-100% higher than in trailing axles

Lead axleTrailing axle

(July 2019)


Model Details► Cut spikes:

• Geometry simplification: removed head• Key geometry details: 6” long by 0.625in. Square• Material properties: validated via lab tensile testing of actual spike

► Screw spikes: • Geometry simplification: Removed threads and head, • Key geometry details: 11/16” diameter threaded shaft• Material properties: same as cut spike (confirmed by manufacturer’s lab testing)

► Single spike timber block:• Key geometry details: 7” by 9” by 18” • Material properties: validated UMAT utilized in previous studies; hourglass control

► Single rail seat model:• Representative system for North American heavy axle load (HAL) Class I railroad• Components: cut spikes, timber block, • Block geometry details: 7” by 9” by 51”• Quantity of spikes: four or five (dependent on study)

► Elements: 8-node linear brick, reduced integration used for all parts except the plate

mechanistic investigation of timber crosstie spike failures · 2020-07-07 · railtec at illinois |...

Documents